Short-Term Power Load Forecasting based on Multi-Factor Similar Day Selection and Multi-Strategy Improved Whale Optimization Algorithm

Short-term power load is susceptible to the meteorological environment and has obvious cyclical changes. To further improve accuracy, a short-term power load forecasting method based on multi-factor similar day selection and multi-strategy improved whale optimization algorithm (MSI-WOA) is proposed....

Full description

Saved in:
Bibliographic Details
Published in:2025 IEEE 3rd International Conference on Power Science and Technology (ICPST) pp. 1860 - 1868
Main Authors: Liu, Yutao, Wu, Guihong, Wang, Xuan
Format: Conference Proceeding
Language:English
Published: IEEE 16.05.2025
Subjects:
Accuracy
Convergence
Forecasting
gated recurrent unit
Heuristic algorithms
Load forecasting
Load modeling
multi-strategy improved WOA
Optimization
Predictive models
short-term load forecasting
similar day
Simulated annealing
Whale optimization algorithms
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Short-term power load is susceptible to the meteorological environment and has obvious cyclical changes. To further improve accuracy, a short-term power load forecasting method based on multi-factor similar day selection and multi-strategy improved whale optimization algorithm (MSI-WOA) is proposed. Firstly, a similar day selection method is established by considering meteorological characteristics and time-cycle factors. Secondly, 16-dimensional input features are constructed, which contain meteorological, time-cycle, and similar day load data. Then, chaotic inverse learning is used to obtain a better initial solution. Nonlinear convergence factors, dynamic inertia weights, adaptive variational perturbations, and simulated annealing strategy are introduced to enhance the global and local search capabilities of the traditional whale optimization algorithm. MSI-WOA is used to optimize the hyperparameters of gated recurrent unit (GRU), such as the number of neuron nodes of two layers, iteration number, and learning rate, to obtain the optimal forecasting model. Finally, after the verification of actual cases and comparative experimental analysis, the proposed method has a better prediction effect.
DOI:10.1109/ICPST65050.2025.11089039